Adaptive Parameters within Genetic Algorithm for Machine Layout Design

Abstract

Finding an optimal solution is always a crucial topic in the field of operations research and management science. In the stochastic search process, the performance of metaheuristics usually depends on the setting of its parameters. A majority of the research in this area is often focused on the static parameter settings adopted from previous research or the best guess approach. In this paper, an adaptive Genetic Algorithm (aGA) is proposed for solving the machine layout design (MLD) problem. In the adaptive process embedded in the aGA, the parameter was dynamically adjusted according to the standard deviation of fitness values during the evolution process. The proposed algorithm was aimed at minimising the total handling distance of materials flowing between non-identical rectangular machines located on the manufacturing shop floor. A series of computional experiments was designed and conducted using five datasets, four of which were adpoted from the literature with another larger dataset generated. Three GA adaptive parameters with three adaptive rates were investigated in the adaptive process, by which the adaptive rate of the mutation operator significantly affected the total material handling distances in the large problem. The statistical analysis of the experimental results suggested that the proposed aGA was able to increase the diversity of chromosomes during the searching process, especially for the largest-size problem. The appropriate adaptive parameters for each dataset were different. The average distances obtained from each problem using the proposed adpative GA parameter setting were significantly lower than those obtianed from the GA with the conventional setting. It was also found that the quality of the best-so-far solutions obtained from the GA with both adaptive and optimised parameter settings were statistically insignificant.